A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 21:00 UTC
[misc] Add FlashMLA as a new option of VLLM_ATTENTION_BACKEND env (#14267)
Serena committed
Mar 6, 2025 at 05:28 UTC
1b7624bf5cfc105a53a5f3f3f5486711ab15d429
1 file changed
+1
vllm/envs.py
+1
@@ -321,6 +321,7 @@ environment_variables: dict[str, Callable[[], Any]] = {
321
# - "XFORMERS": use XFormers
322
# - "ROCM_FLASH": use ROCmFlashAttention
323
# - "FLASHINFER": use flashinfer
324
+ # - "FLASHMLA": use FlashMLA
325
"VLLM_ATTENTION_BACKEND":
326
lambda: os.getenv("VLLM_ATTENTION_BACKEND", None),
327